Error when running inference on custom trained Cascade RCNN

Thanks for your error report and we appreciate it a lot.

Checklist

1. I have searched related issues but cannot get the expected help.
2. The bug has not been fixed in the latest version.

Describe the bug A clear and concise description of what the bug is. I trained a cascade rcnn on custom data successfully. But when I run inference using the trained model I get an error: RuntimeError: The size of tensor a (81) must match the size of tensor b (10) at non-singleton dimension 1

Reproduction

1. What command or script did you run?

result = inference_detector(model, image_path)

1. Did you make any modifications on the code or config? Did you understand what you have modified? Modified number of classes and parameters for custom dataset. The model trained successfully.
2. What dataset did you use? Custom dataset with 9 classes + 1 background class

Environment

1. Please run python tools/collect_env.py to collect necessary environment infomation and paste it here.

sys.platform: linux
Python: 3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
CUDA available: True
CUDA_HOME: /usr/local/cuda-10.1
NVCC: Cuda compilation tools, release 10.1, V10.1.243
GPU 0: Tesla V100-SXM2-16GB
GCC: gcc (Ubuntu 5.4.0-6ubuntu1~16.04.12) 5.4.0 20160609
PyTorch: 1.3.1
PyTorch compiling details: PyTorch built with:
  - GCC 7.3
  - Intel(R) Math Kernel Library Version 2018.0.3 Product Build 20180406 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v0.20.5 (Git Hash 0125f28c61c1f822fd48570b4c1066f96fcb9b2e)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CUDA Runtime 10.1
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_61,code=sm_61;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_37,code=compute_37
  - CuDNN 7.6.3
  - Magma 2.5.1
  - Build settings: BLAS=MKL, BUILD_NAMEDTENSOR=OFF, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -fopenmp -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -O2 -fPIC -Wno-narrowing -Wall -Wextra -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Wno-stringop-overflow, DISABLE_NUMA=1, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=True, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, USE_STATIC_DISPATCH=OFF, 

TorchVision: 0.4.2
OpenCV: 4.1.2
MMCV: 0.2.16
MMDetection: 1.0rc1+78529ec
MMDetection Compiler: GCC 5.4
MMDetection CUDA Compiler: 10.1

1. You may add addition that may be helpful for locating the problem, such as
   • How you installed PyTorch [e.g., pip, conda, source]
   • Other environment variables that may be related (such as $PATH, $LD_LIBRARY_PATH, $PYTHONPATH, etc.)

Error traceback If applicable, paste the error trackback here.

Traceback (most recent call last):
  File "prepare_mmdet_detections_txts.py", line 36, in <module>
    bboxes, labels = image_detections(image_path)
  File "prepare_mmdet_detections_txts.py", line 16, in image_detections
    result = inference_detector(model, image_path)
  File "/home/ubuntu/repos/mmdetection/mmdet/apis/inference.py", line 86, in inference_detector
    result = model(return_loss=False, rescale=True, **data)
  File "/home/ubuntu/anaconda3/envs/pytorch_p36/lib/python3.6/site-packages/torch/nn/modules/module.py", line 541, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/ubuntu/repos/mmdetection/mmdet/core/fp16/decorators.py", line 49, in new_func
    return old_func(*args, **kwargs)
  File "/home/ubuntu/repos/mmdetection/mmdet/models/detectors/base.py", line 140, in forward
    return self.forward_test(img, img_meta, **kwargs)
  File "/home/ubuntu/repos/mmdetection/mmdet/models/detectors/base.py", line 123, in forward_test
    return self.simple_test(imgs[0], img_metas[0], **kwargs)
  File "/home/ubuntu/repos/mmdetection/mmdet/models/detectors/cascade_rcnn.py", line 353, in simple_test
    cls_score = sum(ms_scores) / self.num_stages
RuntimeError: The size of tensor a (81) must match the size of tensor b (10) at non-singleton dimension 1

Bug fix If you have already identified the reason, you can provide the information here. If you are willing to create a PR to fix it, please also leave a comment here and that would be much appreciated!

[shliang0603]

I meet same error, I have solved it.

You should modify all num_classes for yourself classes number. Some model configuration files have several num_classes modifications

[changruowang]

Maybe your three num_classes is not same